Medicine is shifting toward a predictive, participatory, and proactive mode. A logical extension of this shift is the use of patient-reported health-related quality of life (HRQL), together with other biomedical information, to predict future health. More accurate prediction of future health states will allow for better decision making and more informed and timely interventions. Over the past decade, research into the prognostic value of patient-reported outcomes, including HRQL, has found independent associations between baseline assessments and survival in cancer and other diseases. But longitudinally-observed HRQL may be more relevant in complex medical conditions and treatments such as hematopoietic stem cell transplant (HSCT). For more than 40 years, HSCT has offered life-saving therapy to children and adults with high risk malignancies and other life-threatening diseases of hematologic, immunologic, and metabolic origin. HSCT is now the second most common form of transplantation in the US. Although transplant predominantly occurs at regional facilities, subsequent care is now shared with local providers. At the center of that care is the recipient and, in the case of children, his/her parent(s) who serves the complex role of comforter, treatment enforcer, and care coordinator. Currently there is no conduit for the child's and parent's perception of the evolving HRQL trajectory to inform clinical decision making. For example, the functional decline in a child whose status appears stable by routine clinical indicators, may serve as an early warning signal of an insipient clinical event (e.g., chronic graft versus host disease or infection). This child could benefit from rapid transfer back to the transplant center for aggressive evaluation and treatment. One of the barriers to the incorporation of longitudinal HRQL information in the management of complex medical conditions is the lack of statistical methodology for the integration of longitudinal outcomes with multiple sequential and simultaneous clinical outcomes. Standard survival analyses with time-varying HRQL as a covariate would not be valid because HRQL trajectories are affected by many of the same factors that affect the clinical course. Joint models for longitudinal HRQL and clinical outcomes are needed. Although such models have been developed for a single clinical outcome, no models exist that can incorporate HRQL with a multi-state process that includes intermediate states, such as toxicity from treatment, and terminal states, such as recovery or mortality. We will address this gap by developing new statistical methodology for joint models for longitudinal outcomes and multi-state processes, with application to HRQL in pediatric HSCT. We will give clinical researchers the tools to assess the value of longitudinal HRQL data for improving prediction of sequential and simultaneous clinical events in complex medical conditions. This project will take an essential step toward enabling patients, families, and providers to utilize changing HRQL information to facilitate communication and enhance clinical decision-making, ultimately leading to better patient outcomes. PUBLIC HEALTH RELEVANCE: Medicine is shifting toward a predictive, participatory, and proactive mode. A logical extension of this shift is the use of patient-reported quality of life, together with other biomedical information, to predict future health. This project will give clinical researchers the statistical tools to assess the value of quality of life trajectories for predicting patient outcomes in complex medical conditions, thus taking an essential step toward enabling patients, families, and providers to utilize changing quality of life information to facilitate communication and enhance clinical decision-making.